Author Affiliations
Abstract
1 Aerospace Laser Technology and Systems Department, Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai 201800, China
2 Center of Materials Science and Optoelectronics Engineering, University of Chinese Academy of Sciences, Beijing 100049, China
3 Shanghai Typhoon Institute, China Meteorological Administration, Shanghai 200030, China
A method of spectrum estimation based on the genetic simulated annealing (GSA) algorithm is proposed, which is applied to retrieve the three-dimensional wind field of typhoon Nangka observed by our research group. Compared to the genetic algorithm (GA), the GSA algorithm not only extends the detection range and guarantees the accuracy of retrieval results but also demonstrates a faster retrieval speed. Experimental results indicate that both the GA and GSA algorithms can enhance the detection range by 35% more than the least squares method. However, the convergence speed of the GSA algorithm is 17 times faster than that of the GA, which is more beneficial for real-time data processing.
coherent Doppler lidar three-dimensional wind field retrieval genetic simulated annealing algorithm spectrum estimation typhoon Chinese Optics Letters
2024, 22(4): 040101
1 巢湖学院电子工程学院， 安徽 合肥238000
2 安徽省气象台，安徽 合肥230031
风场对于天气形势的演变和预报至关重要。基于风云四号A星干涉式大气垂直探测仪（GIIRS）中波通道资料和ERA5风场资料，采用LightGBM进行大气三维风场反演研究。首先，构建模型特征变量。GIIRS通道最优选择采用二步特征选择法：（1）建立GIIRS通道黑名单；（2）采用置换特征重要性（Permutation Feature Importance，PFI）方法选择特征变量，在形成通道最优子集的基础上，构建含有时空信息的特征变量。其次，构建基于LightGBM的三维风场反演方法。最后，基于台风“利奇马”期间的GIIRS加密资料开展了LightGBM超参数优化和相关反演试验。结果表明，相对于ERA5风场资料，测试集中风场U和V分量的均方根误差（Root Mean Square Error，RMSE）分别小于1 m/s和15 m/s。本文中的二步特征选择法能够实现GIIRS通道的动态最优选择。
大气风场反演 特征选择 台风“利奇马” FY-4A/GIIRS FY-4A/GIIRS atmospheric wind field retrieval feature selection LightGBM LightGBM Typhoon “Lekima”
1 西南技术物理研究所,成都 610041
2 成都理工大学 信息管理学院,成都 610059
3 中国人民解放军 63891部队,洛阳 471003
为了提高多普勒激光测风雷达反演风场的速度和精度,采用Newton-Gaussian最小二乘优化方法对风场反演模型进行数值求解,并对Newton-Gaussian法的步长和初值进行讨论和调整,使得此算法的鲁棒性更强,同时也降低了传统速度方位显示反演方法中固有缺陷的影响。将其应用到研制的全光纤多普勒激光测风雷达系统中,实现了对低中空局部风场的反演。结果表明,该算法不但提高了风场反演的速度和精度,而且反演结果更稳定,适合实时计算。
激光技术 风场反演 速度方位显示法 Newton-Gaussian法 多普勒测风激光雷达 laser technique wind field retrieval veloaity azimuth display method Newton-Gaussian method Doppler wind lidar
